Batching `reflect_table` for improved performance

[zzzeek]

Discussed in #1766

I'm sure this was brought up elsewhere as well, it's time to ensure that alembic autogenerate makes use of SQLAlchemy's batched reflection API so that autogenerates can be as much as 10x faster on some platforms.

[sqla-tester]

Mike Bayer has proposed a fix for this issue in the main branch:

WIP: new pre-caching architecture for autogenerate https://gerrit.sqlalchemy.org/c/sqlalchemy/alembic/+/6531

[sqla-tester]

Mike Bayer has proposed a fix for this issue in the main branch:

new pre-caching architecture for autogenerate https://gerrit.sqlalchemy.org/c/sqlalchemy/alembic/+/6541